I e-mailed them to ask but they wouldn't tell me out of privacy concerns. If I remember correctly they said it was on the Ionian side. I should probably e-mail them again and tell them how strangely this sample behaves, and that they might be using a highly drifted and not very representative sample for South Italians.
I've been complaining about the samples used for some of these countries for years, not that anyone cares. In addition to the actually Thessalonian sample used for Greece (which is why it plots so close to Toscana), there's the half Catalan sample used for Iberia. If you listen to the Catalans, they're in no way Iberians, and that includes genetically!
Hold on. I see that there's no Iberian sample in this particular plot, only one labeled Catalan. If it's the oft used academic sample, it's actually half Catalan. Generally, it's pretty clear to me that Iberians list north not because of additional "Celtic", given that they have so little "steppe" ancestry, but because they absorbed a bit more WHG than people in the southeast.
These researchers seem to have blind spots that won't be dislodged by "amateurs" e-mailing them. Take, for example, the Hellenthal/Busby supposed analysis of "recent" dna absorption. I think a lot of it is wrong or at least questionable. For example, I'm not sure they can tell if there was an inflow of more southeastern, Greek like (farmer heavy) dna into more central Euro type people in Italy after the fall of Rome, or the opposite, i.e. "Celtic", "Lombard" admixture into a much more "Mediterranean" population, which they date too late because as is always the case with the program they use, it picks up only the date of the latest admixture.